CORPUSAI citability
What is llms.txt? Honest data on who reads it.
Corpus Plan generates llms.txt because it is cheap, clear housekeeping for a site that wants to be easier for agents to inspect. We do not sell it as a ranking factor, citation trigger, or AI-visibility promise.
Every outside claim on this page has a numbered source. Accessed 2026-09-02.
The 28% and 97% figures are from Ahrefs' 137K-domain analytics sample, not the whole web.[2]
The plain definition
/llms.txt is a proposed Markdown file at the root of a website. The proposal describes it as a way to give agents a concise guide to the useful parts of a site, usually with a title, short summary, and links to important pages or Markdown versions of pages.[1]
For a small business, that usually means a simple file with the business name, what the business does, location or service area, primary service pages, contact page, and any pages that answer customer questions.
What the data says
The adoption number is real enough to notice: Ahrefs reported that 28% of the 137K domains using its web analytics had an llms.txt file. But readership was thin. In the same study, 97% of those llms.txt files received zero requests in May 2026.[2]
That is the reason Corpus Plan treats the file as a parity fix, not the center of AI search work. Publishing it may help tools and agents that look for it. The available data does not support telling a business that the file will make ChatGPT, Claude, Perplexity, or Google cite them.
Who does read it?
The most defensible answer is narrow: some tools, documentation platforms, coding agents, auditors, and bots. The v2 change notes for the proposal point to stronger use by documentation platforms and coding agents than the original 2024 proposal could assume.[3]
That still leaves a gap between "some agents can use this" and "AI answer engines rely on this for mentions." Corpus Plan only claims the first.
Should a small business publish one?
Yes, if it takes minutes and does not distract from higher-value fixes. A clean llms.txt is like a sitemap with short notes for agents: inexpensive, readable, and easy to check. It should sit behind schema, indexable content, good titles, clear service pages, contact details, and crawler access in the priority list.
Do not buy a big project around it. Do not treat it as proof that AI assistants understand your business. Do publish it when the rest of the site already says who you are in plain, crawlable HTML.
A safe small-business template
# Business Name
> One sentence describing what the business does, who it serves, and where.
## Key pages
- [Services](https://example.com/services): Main services and who they are for.
- [Service area](https://example.com/service-area): Cities or regions served.
- [About](https://example.com/about): Ownership, credentials, and business background.
- [Contact](https://example.com/contact): Phone, email, address, hours, and booking path.
- [FAQ](https://example.com/faq): Direct answers to common buyer questions.Keep it factual. Do not stuff it with keywords, claims, or promises your site does not support.
How Corpus Plan scores it
The free audit checks whether /llms.txt exists and can be fetched. Missing llms.txt is a minor GEO finding, not a critical failure. The report explains the tradeoff in the same terms as this article: proposed standard, low confirmed AI-crawler use today, cheap to publish.
Pro can generate the file from your site details, but the promise is mechanical: a clean file you can place on your site. It is not a promise of rankings, traffic, or AI mentions.
Sources
References
- https://llmstxt.org/ — accessed 2026-09-02. The llms.txt proposal: a Markdown file at /llms.txt that helps agents find the useful parts of a site.
- https://ahrefs.com/blog/llmstxt-study/ — accessed 2026-09-02. Ahrefs usage study: 28% of 137K domains in its analytics sample published llms.txt, while 97% of those files received zero requests in May 2026.
- https://llmstxt.org/changes.html — accessed 2026-09-02. llms.txt v2 change notes: adoption by documentation platforms and reliable use by coding agents, with the original proposal still described as speculative when published.
Spotted an outdated source or a claim that needs correction? Contact Corpus Plan and we will re-check it.